Oil & Gas Science and Technology Rev. IFP, Vol. 54 (1999), No. 6, pp.

759-771
Copyright 1999, ditions Technip

Constitutive Laws for Geomaterials

E. Papamichos1
1 SINTEF Petroleum Research, N-7465 Trondheim - Norway
e-mail: euripides.papamichos@iker.sintef.no

Rsum Lois constitutives pour les gomatriaux Il existe un nombre important de lois
constitutives permettant d'approcher le comportement des gomatriaux sous diverses contraintes,
conditions aux limites et divers chemins de chargement, pressions de pore, tempratures, etc.
Llaboration de lois constitutives suit deux tendances concurrentes : la tendance laborer une loi
unifie dcrivant le comportement du matriau sous les conditions les plus gnrales, et le besoin de lois
justifi par une utilisation efficace dans la pratique. La dernire tendance inclut la fois la possibilit
dune calibration par des essais sur des chantillons disponibles, et la convivialit de la loi. des fins
dingnierie, la seconde tendance domine, lingnieur doit valuer le problme et slectionner le modle
le plus appropri pour dcrire le phnomne dominant observ. Dans cet article, un examen des diverses
lois constitutives pour les gomatriaux, et en particulier pour leurs applications, est prsent. Divers
exemples de problmes concernant lingnierie et les gomatriaux appliqus au ptrole sont utiliss afin
de dmontrer que la loi doit tre aussi simple que possible, mais pas trop non plus.
Mots-cls : loi constitutive, grs, craie, injection deau, endommagement.

Abstract Constitutive Laws for Geomaterials An abundance of constitutive laws exists to

approximate the behavior of geomaterials under various stresses, boundary conditions and loading
paths, pore pressures, temperatures and so on. The construction of constitutive laws is driven by two
competing trends: the tendency for a unifying law describing the material behavior under the most
general conditions, and the need for laws that can be used efficiently in practice. The latter incorporates
both the possibility of calibration from available specimens and the user-friendliness of the law. For
engineering purposes, the second tendency dominates and the engineer scientist has to evaluate the
problem and select the most appropriate model to describe the dominant phenomena on hand. A review
of various constitutive laws for geomaterials is presented with an emphasis on their application. Various
examples of engineering problems and geomaterials with emphasis on petroleum applications are used
to demonstrate that the law must be as simple as possible but not simpler.
Keywords: constitutive law, sandstone, chalk, water injection, core damage.

INTRODUCTION continuous, denying the existence of molecules. This is not

The constitutive laws that are presented describe the behavior
of a geomaterial as a continuum. The perspective of the
continuum material description can be appreciated in the
following quote by Truesdell and Noll (1965):
Widespread is the misconception that those who
formulate continuum theories believe matter really is
so. Continuum physics presumes nothing regarding the
structure of the matter. It confines itself to relations among
gross phenomena, neglecting the structure of a material on
a smaller scale. Whether the continuum approach is
justified, in any particular case, is a matter, not for the
philosophy or methodology of science, but for the
experimental test
760 Oil & Gas Science and Technology Rev. IFP, Vol. 54 (1999), No. 6
Continuum models are indeed widely used today for both
research and engineering applications and provide useful
solutions to technical problems. This of course is not to say
that micromechanical and discontinuum models should be
disregarded as purely scientific in nature. Without a doubt
such models have shown a significant potential and in some
cases they constitute the preferred way of modeling.
The literature on constitutive laws for geomaterials is rich
in models. The reason is that a wide variety of geomaterials
with often very dissimilar behaviors has to be appropriately
described, such as the behavior of a sandstone, chalk, sand,
limestone, shale, concrete, etc. In addition to the various
geomaterials, the behavior of a particular material may
depend significantly on the application at hand, that is on the
boundary or loading conditions, the pore pressures, the pore
fluids, the scale of the problem, the time periods involved,
the temperature, etc.
There are two trends in constitutive modeling. The first is
to make the model as general as possible in the expense of
complexity and the other to make it as problem-specific as
possible with the advantage of simplicity. The characteristics
of a general constitutive model are:
Various material types are included, such as elastic, plastic,
creep, pore pressure, temperature, water saturation, etc.
Various problems can be solved. The model can be used in
various different problems such as wellbore stability, sand
production, hydraulic fracturing, compaction, reservoir
modeling, etc.
The role of the engineer is limited.
Extensive calibration is needed. The model requires the
input of many material parameters, which are often
difficult to obtain, as in the case of core material where
enough test specimens are not available, or expensive to
obtain, as large test series are required.
Low efficiency. The model is cumbersome to use and often
the results are difficult to interpret, as the important
physical mechanisms are not always obvious.
On the other hand the characteristics of a problem-specific
constitutive model are:
Limited number of material types are included, such as
elastic or elastic-plastic or creep or pore pressure or
temperature or water saturation, etc.
Limited number of problems can be solved, as for example
wellbore stability or sand production or hydraulic
fracturing or compaction or reservoir modeling, etc.
The role of the engineer is significant since evaluation of
the geomaterial and problem at hand and selection of the
appropriate model are required.
Limited calibration is needed, as only the absolutely
necessary material parameters are required.
Optimum efficiency is attained as non-important effects
are excluded.
For engineering problems the problem-specific constitutive
models have obvious advantages. Examples of such problems
with emphasis on petroleum applications are given to
demonstrate how to select or develop a constitutive model
based on experimental and/or field data and observations.
Section 1 describes a sandstone model for open hole stability
and sand production problems, Section 2 a chalk model for
reservoir compaction with water saturation effects, and
Section 3 a sandstone model for core damage effects. Finally
we present the concluding remarks.
1 SANDSTONE MODEL FOR OPEN HOLE STABILITY
AND SAND PRODUCTION
Open hole stability and sand production problems are often
encountered in sandstone reservoirs during hydrocarbon
production. The reservoir rocks are usually soft, weak
sandstones of considerable porosity in the range of 10-30%.
In this problem, the rock is exposed to relatively low mean
stresses by high deviatoric stresses due to the stress
concentration around the hole.
A nonlinear elastic-plastic model has been developed to
describe the near well behavior of these sandstones (Sulem et
al., 1999). In brief, the model has an elasticity, which is
nonlinearly stress-dependent, and a plasticity based on the
flow theory. The plasticity is related to a linear Mohr-
Coulomb yield function F with friction hardening and
cohesion softening. The plastic dilatancy is described by a
linear Mohr-Coulomb plastic potential function, different
from the yield surface, which makes the flow rule not
associative. This allows for the description of both
compactive and dilatant behaviors under shear straining. On
the other hand, the appropriate plastic behavior under
hydrostatic straining has not been incorporated in the present
model as plastic hydrostatic straining is not usually
significant around wellbores and perforations. Throughout
this section compression is taken negative, as it is more
convenient for calculations. Tests results and model
predictions are given with compression positive as more
often used in rock and soil mechanics.
1.1 Model Description
The development of the model is based on behavior of soft,
weak sandstones as observed in triaxial compression and
extension experiments (Papamichos et al., 1999), which are
relative experiments for obtaining the deviatoric behavior of
a material. For nonlinear materials, the constitutive relations
are formulated in the incremental form which, for time-
independent behavior, is equivalent to the rate form
employed in the following. A dot over a quantity denotes the
rate (or increment) of the quantity.
 E Papamichos / Constitutive Laws for Geomaterials 761

The rate form of the constitutive stress-strain relationships where I1 is the first stress invariant and J2, the second
can be written as: deviatoric stress invariant. The dependency of G on the
equivalent stress e, is described by the material function:
ij = Cijkl
ep
kl (1)
(G2 + G3 e ) e
where ij and ij are the stress and strain rates, respec- G0 + for e epeak
G = G( e ) = 1 + G1 e (6)
ep
tively, and Cijkl the elastic-plastic tangent stiffness tensor, G for e > peak
peak e
which can be written as:
where the constants G0 to G3, Gpeak, and epeak for the various
ep
Cijkl = Cijkl
e
Cijkl
p
(2) materials are obtained by fitting this function to the
Figure 1 shows triaxial-compression test results on humid experimental data. According to Equation (6), the elastic
Red Wildmoor sandstone at three confining stresses. Starting shear modulus G increases from the initial value G = G0 at
zero stress to G = Gpeak at e = e and remains constant
peak
from the elasticity of the material, the load-unload cycles
e
show that the elasticity is nonlinear and depends on the thereafter. The elastic tangent stiffness tensor Cijkl may then
stress. In order to keep the model as simple as possible, it is be written as:
assumed that the Poissons ratio is constant and thus the 2
elastic shear modulus G must be a function of the stress, that
e
Cijkl = G ij kl + ik jl + il jk
is G = G(ij). The reversibility and path independency of the 1 2
(7)
elastic strain energy requires that the stresses are obtained ij kl dG d e
from the strain energy function W as (Chen and Han, 1988): + il jk + 2(1 + )
e2 G e dG d e
W
ij = (3) With respect to plasticity, the standard flow theory is used
ij p
where the plastic tangent stiffness tensor Cijkl is expressed as:
With these assumptions and requirements, the elastic
shear modulus G must be a function of the equivalent stress 1 e Q F e
p
Cijkl = Cijmn Cstkl (8)
e, defined in terms of the isotropic (mean) stress p and the h mn st
shear-stress intensity as:
with F and Q being the yield and plastic potential functions,
e = 3(1 2 ) p 2 + 2(1 + ) 2 (4) respectively. The hardening modulus h is given by:

The stress invariants p and are defined as: F e Q F d F dq

h= Cijkl Qs + (9)
p = I1 3, = J2 (5) ij kl dg p
q dg p

40 -8 Confining stress 0.5 MPa

35
Confining stress 6 MPa -6
Volumetric strain (x 10-3)

30
Axial stress (MPa)

25 Confining stress 2.5 MPa -4

20 -2 Confining stress 2.5 MPa

15 Confining stress 0.5 MPa 5 10 15 20 25

0
0
10

5 2
Confining stress 6 MPa
0 4
0 5 10 15 20 25
Axial strain (x 10-3)
a Axial strain (x 10-3) Red Wildmoor b
Figure 1
Triaxial compression test results on humid Red Wildmoor sandstone. a: axial stress, and b: volumetric strain versus axial strain at three
confining stresses (Papamichos et al., 1999).
762 Oil & Gas Science and Technology Rev. IFP, Vol. 54 (1999), No. 6

and the following definition of the switch function ij and the accumulated plastic shear-strain intensity gp. The
1 expressing the plastic loading criterion: use of nonassociativity where the yield and plastic potential
F e functions do not coincide is necessary in soft weak

1, for F = 0 and Cijkl kl > 0
sandstones. The experimental results in Figure 1b show that
ij
the material initially compacts and then dilates as the peak
1 = (10)
0, for F < 0 or F = 0 and F C e 0 stress is approached. With associative behavior, dilation is

ij
ijkl kl
always predicted. For the Mohr-Coulomb criterion, F and Q
can be written in terms of the stress invariants p, , and , as:
Figure 2 shows the peak axial stress during triaxial
sin
compression at various confining stresses, which gives the
shape of yield surface at peak stress. The yield surface is ( )
F = F ij , g p = cos
3
( q p) = 0
approximated with a straight line in order not to include (11)
sin
additional nonlinearities in the model, such as for example a (
Q = Q ij , g p
) = cos
3
+ p
pressure dependency of the strength parameters. At higher
confining stresses the curvature of the yield surface may be where the stress invariant is called the Lode angle and
large enough, such that the exclusion of pressure-dependent describes the relative magnitude of the intermediate principal
effects on the strength parameters may not be possible. For stress. It is defined through the second J2 and third J3
the selection of the yield surface, reduced triaxial extension deviatoric stress invariants as:
test data suggest that the material follows an intermediate
behavior between the Mohr-Coulomb and the Drucker- 1 3 3 J3
= a sin 32
(12)
Prager yield surfaces (Papamichos et al., 1999). Other 3 2 J2
analyses and comparisons between model predictions and
test results suggest that the Mohr-Coulomb model is more The parameter q = q(gp) is the tension cut-off, and
appropriate for geomaterials. The Mohr-Coulomb is a model = (gp) and = (gp) are the mobilized friction and
widely used in geomechanics and a lot of experience and dilatancy coefficients, respectively. These are related to the
strength parameters, such as cohesion and friction angle, are mobilized friction angle m, dilation angle m, and cohesion
available. c through the expressions:
Figure 1a shows that the material behaves plastically = sin m , = sin m , q = c cos m (13)
almost from the start of straining as demonstrated by the
residual plastic strains during the unload-reload cycles. The parameter Qs in Equation (9) is related to the plastic
During the stress increase, the material hardens with the potential surface Q and it is given as a function of the stress
accumulated plastic shear-strain intensity gp, and after the invariants J2, J3 as:
peak stress is reached, the material softens. The yield F and 2 2

plastic potential Q surfaces are then functions of the stress J 2 B1Q + 3 J3 B1Q B2Q + J 22 B2Q (14)
Qs = 2
3
where:
60 cos 1 + tan tan 3 + ( tan 3 tan )

50 2 3
Q
3 +
Axial stress (MPa)

40 B1Q = = if =
J 2 4 3 6
30 3 (15)
if =
4 3 6
20
3 sin + cos
10 Red Wildmoor sandstone

B2Q =
Q
= 2 2 cos 3
0 J3
0 if =
0 2 4 6 8 10 12 6
Confining stress (MPa)
Finally, the plastic strain-rates ij are given by the flow
p
Figure 2
Triaxial compression test results on humid Red Wildmoor rules:
sandstone. Peak axial stress versus confining stress and
straight line fit to the data (Papamichos et al., 1999). Q 1 F e
ijp = , = Cijkl kl (16)
ij h ij
 E Papamichos / Constitutive Laws for Geomaterials 763

which result into the following expression for the plastic The model can be calibrated through a minimum of two
shear-strain intensity rate gp : triaxial compression tests with load-unload cycles at various
gp = Qs (17) stress levels. Table 1 lists values of the various elastic
and plastic material constants for humid Red Wildmoor
For the elastic strain rates it holds that:
p
ij sandstone.
ije = ij ijp (18) TABLE 1
The model requires the calibration of the material Humid Red Wildmoor sandstone: elastic and plastic material constants
functions q = q(gp), = (gp) and = (gp). For these
sandstones, a model with both hardening and softening with [] 0.2 f0 [ ] 0.067148 q0 [MPa] 3.82
plastic shear strain was used, as illustrated in Figure 3.
Hardening was modeled by an increasing friction coefficient , G0 [MPa] 349.17 f1 [ ] 109.88 q1 [MPa] 25.0
which reaches its maximum value at plastic shear strain G1 [MPa-1] 0.042907 f2 [ ] 273.75
g p = gpeak
p
and remains constant thereafter. Softening was
modeled by a tension cut-off q, which starts to decrease from G2 [ - ] 250.32 f3 [ ] 8946.6 d1 [ ] 5.834
its initial value q0 after the friction coefficient has reached its G3 [MPa-1] 0.0 fpeak [ ] 0.94523 d2 [ ] 0.78636
peak value, and thus after the plastic shear strain gp has
p p
reached the value gpeak . In particular, the following functions Gpeak [MPa] 6183.2 gpeak [ ] 0.0099069 d3 [ ] 90.453
were calibrated for each material:
epeak [MPa] dpeak [ - ] 0.94523
q0 for g g
p p
peak
q= p p
(
q0 q1 g gpeak ) for g p > gpeak
p
1.2 Calculation Examples

f0 +
f2 + f3 g g ( p
) p

for g p gpeak
p The back analysis of the triaxial compression and reduced
3f
= , f = 1 + f1 g p extension test results is shown in Figure 4. The comparison
2 3+ f with the experimental results for both test types is very
fpeak for g p > gpeak
p
(19)
satisfactory despite the complexity of the soft weak
d2 + d3 g p 2 + d1g p

(
for g p gpeak
p ) sandstone behavior. The model was also used to predict the
=
3d
2 3+d
, d= 1 + d1 g p
( ) response of Red Wildmoor sandstone hollow cylinders
loaded isotropically (Papamichos et al., 1996). In these tests

d peak for g p > gpeak
p
a thick-wall hollow cylinder was loaded externally, while the
internal hole remained unsupported. The external radial stress
p
where the constants q0, q1, gpeak , f0 to f3, fpeak, and d1 to d3, was equal to the axial stress. The internal hole and external
dpeak are obtained by fitting the functions (19) on experi- deformations of the cylinder were measured. In Figure 5
mental data. experimental results for the external radial stress versus the
internal and external tangential strains are compared with the
model predictions. The nonlinear elastic-plastic model
fp predicts very well the hollow cylinder response up to the
1 (1) T point of failure of the inner hole. This is expected since after
this point the continuum description of the material ceases.
cp
The point of initial failure is predicted through the
f0 bifurcation theory and continua with microstructure. In
1 (2) addition to the nonlinear elastic-plastic model, predictions
(0) c were made using a simpler model where the nonlinear
elasticity was suppressed and an average value of elastic
c0 shear modulus was used. It can be seen in Figure 5 that the
linear elastic-plastic model predictions are rather poor,
which illustrates the need for a nonlinear elasticity model
q q0 p in this case.
For weakly cemented, unconsolidated sandstones the
Figure 3
constitutive law may need to be enriched with additional
Motion of the Mohr-Coulomb yield surface in stress-space: features to model adequately the near well behavior of
(0-1) friction-hardening phase, (1-2) tension-softening phase
the rock. In the presented model, a linear Mohr-Coulomb
(Sulem et al., 1999).
model was adopted for relatively low mean stresses. In
764 Oil & Gas Science and Technology Rev. IFP, Vol. 54 (1999), No. 6

Simulation of triaxial compression tests 60

Simulation of triaxial
60 (complete model) extension tests
Data Computed (complete model)
Data
10.3 MPa

Axial stress (MPa)

40 Computed
Axial stress (MPa)

40

6.9 MPa
3.5 MPa

20 20

1.4 MPa
0.4 MPa 60 MPa
52.5 MPa
0 45 MPa
0
-0.015 0.000 0.015 0.030 -0.015 0.000 0.015
a Radial strain Axial strain b Radial strain Axial strain

Figure 4
Back analysis of a: triaxial compression, and b: reduced triaxial extension test results (Sulem et al., 1999).

35 30
Simulation
30 Linear elastic-plastic
External radial stress (MPa)
External radial stress (MPa)

25
simulation
25 Test 21Cav04
20
20
Initial failure 15
15
Simulation 10
10
Linear elastic-plastic
simulation 5
5
Test 21Cav04
0 0
0 0.005 0.01 0.015 0.02 0.025 0.03 0 0.0005 0.001 0.0015 0.002 0.0025
a Internal tangential strain b External tangential strain

Figure 5
Hollow cylinder stability. Nonlinear elastic-plastic versus linear elastic-plastic model simulations and comparison with experimental data.

unconsolidated sandstones, however, a curved yield surface
may be required even at low mean stresses. This may be
modeled with a pressure-sensitive friction angle. The
addition of a pressure cap may also be necessary. Another
observation in these sandstones is their volumetric response
upon shearing. Figure 6 shows the volumetric response of an
unconsolidated reservoir sandstone tested in triaxial
compression at three confining stresses. The experimental
results clearly indicate a pressure-sensitive dilatancy, which
may be necessary to be included in the constitutive law.
2 CHALK MODEL FOR RESERVOIR COMPACTION
DUE TO WATER INJECTION
The mechanical behavior of weak, high-porosity chalks is
important in many aspects of oil production such as reservoir
compaction and solids production. Among several factors
affecting the mechanical behavior of such rocks is the relative
fluid saturation (gas/oil/water), which has an impact on the
capillary forces between the grains. Capillarity arises from a
fluid property known as surface tension, a phenomenon
 E Papamichos / Constitutive Laws for Geomaterials
-0.03
Atlas 05 Confining stress 0.5 MPa
-0.02
Volumetric strain
-0.01
0.06
00 0.02 0.04 0.08 0.1 0.12
Atlas 01 Confining stress 2 MPa
0.01
0.02
Atlas 04 Confining stress 10 MPa
0.03
0.04
Axial strain
Figure 6
Triaxial compression test results on an unconsolidated
reservoir sandstone. Volumetric strain versus axial strain at
three confining stresses, showing pressure-sensitive
dilatancy.
that occurs at the interface between different materials. For
rocks and soils it occurs between surfaces of water (normally
the wetting phase), mineral grains, and air, gas or oil
(normally the nonwetting phase). At low degrees of
saturation, water is bound in small pores and in the narrow
spaces between grains forming capillary menisci. Low
degrees of water saturation are often representative of the
initial conditions in oil or gas reservoirs. It is noted that for
some rocks, bridges of connate water are known to be stable
for temperatures up to 300C, posing problems in defining a
dry reference state. The surface tension at the boundary
between the water and the air in the adjoining voids pulls the
grains together with a force known as contact pressure. The
frictional resistance produced by the contact pressure has the
same effect as if the grains were held with a certain amount of
cohesion. At higher degrees of saturation, the voids between
the grains are filled with water, the surface tension strongly
decreases and the contact pressure vanishes. The magnitude
of the capillary suction can reach the order of several MPa,
depending on the radius of the capillary menisci, which is a
function of the pore size. Another characteristic of chalks is
their collapsible behavior. Chalks have a high porosity
because they comprise of coccoliths, that is, skeletons of
marine animals. In some chalks, it is found that capillary
menisci formed by connate water are the only cohesional
force that holds together the coccoliths. Upon elimination of
the cohesional forces between the coccoliths, the chalk
structure collapses with a significant reduction in porosity.
Similar collapsible materials are the bulking structures
observed in unconsolidated sands and sandstones.
An elastic-plastic model with water effects was developed
to model the behavior of partially water-saturated chalks
765
(Papamichos et al., 1997). The model is based on experi-
mental evidence from tests on Pietra Leccese chalk and
incorporates Bishops effective stress principle for the
capillary pressure, a water-saturation-dependent elastic
stiffness, and a Mohr-Coulomb yield surface with a pressure
cap. The cohesion and pressure cap parameters in the yield
surface are water-saturation-dependent to model the capillary
effect on the strength parameters. Compression is taken
negative in this section.
2.1 Experimental Evidence
The model is based on experimental evidence from triaxial
and hydrostatic compression tests on Pietra Leccese chalk at
various degrees of water saturation (Papamichos et al., 1997).
The Pietra Leccese chalk is a high-porosity outcrop bio-
calcarenite from the south of Italy. Figure 7 shows uniaxial
compression test results for specimens at various water
saturations. The results show that the uniaxial compressive
strength (UCS) decreases substantially with increasing water
saturation. Load-unload cycles in these tests showed that for
this material, the elasticity does not change significantly with
the applied axial or confining stress. However, a significant
decrease in the Youngs modulus with increasing water
saturation was observed. The Poissons ratio on the other
hand did not appear to be influenced significantly by the
saturation. From the test data, an average elastic Youngs
modulus E for each saturation, and a constant average
Poissons ratio for all saturations were calculated.
In Figure 8, the peak axial stress obtained at uniaxial and
triaxial compression tests is plotted versus the confining
stress and saturation. From this plot and assuming a linear
Mohr-Coulomb yield criterion for the peak axial stress, the
-50
-45
-40
Axial stress (MPa)
-35
-30
-25
-20
Plt05 S = 0.016
-15 Plt33 S = 0.025
Plt09 S = 0.047
-10
Plt30 S = 0.055
-5 Plt10 S = 1
0
0 -0.001 -0.002 -0.003 -0.004
Axial strain
Figure 7
Uniaxial compression test results on Pietra Leccese chalk.
Axial stress versus axial strain for various water saturations S.
766 Oil & Gas Science and Technology Rev. IFP, Vol. 54 (1999), No. 6
-70
-60
Axial stress (MPa)
-50
-40
-30
-20
-10
0
0 -2 -4 -6 -8
Confining stress (MPa)
Figure 8
Uniaxial and triaxial compression test results on Pietra
Leccese chalk. Peak axial stress versus confining stress and
saturation S.
peak friction angle and cohesion c can be calculated for the
two saturations where triaxial test data are available. The
calculated values of peak friction angle and cohesion are
tabulated in Table 2. From these limited test results, it
appears that changes in the water saturation affect primarily
the cohesion of the material, which reduces substantially in
fully water-saturated specimens. This is attributed to the
destruction, with increasing water saturation, of the capillary
menisci, which act as cohesional bonds between the grains.
On the other hand, the internal friction angle, which is
primarily influenced by such factors as porosity, particle
shape and size, and grain size distribution, appears not to be
affected significantly.
TABLE 2
Pietra Leccese chalk: peak friction angle and cohesion
at various water saturations calculated from triaxial compression test data
Saturation Peak friction angle () Peak cohesion (MPa)
0.047 29.1
1 27.5
Figure 9 shows the results from four drained hydrostatic
compression tests performed on specimens at different water
saturations, ranging from 0.013 to 1. During plastic yield
under hydrostatic compression, the specimens undergo a
characteristic structural change for collapsible materials
which is referred to as pore collapse. This gradual structural
change is characterized with a significant increase in
compressibility. The results show that the elastic bulk
modulus and the plastic yield stress decrease significantly
with increasing water saturation. The tangent bulk modulus
does not change substantially until plastic yield and thus the
elasticity of Pietra Leccese can be considered at a first
Plw10 S = 0.013
-180 Plw22 S = 0.019
Plw23 S = 0.047
-160 Plw31 S = 1
Hydrostatic stress (MPa)
-140
-120
-100
S=1
S = 0.055 -80
S = 0.047 -60
S = 0.025
S = 0.016 -40
-20 Pore collapse
-10 -12 0
0 -0.05 -0.1 -0.15 -0.2
Volumetric strain
Figure 9
Hydrostatic compression tests on Pietra Leccese chalk showing
the influence of water saturation on the pore collapse stress.
approximation as stress-independent. The post-pore-collapse
tangent bulk modulus is approximately the same in all tested
specimens, and thus it can be considered as saturation-
independent.
Three water-injection tests were also performed. The
water-injection tests comprise of three phases. In the first
phase, a specimen with initial saturation 0.013 or 0.019 is
hydrostatically compressed under drained conditions until a
prescribed hydrostatic stress level is reached. In the second
phase, the hydrostatic stress is maintained constant, and water
injection is performed by flowing few pore volumes of
equilibrium water through the specimen at a constant flow
rate. The flow rate is kept low to ensure no significant pore
pressure built-up in the specimen. In the third phase, fluid
flow is halted and the specimen is hydrostatically unloaded
under drained conditions to zero stress. Figure 10 shows plots
of the experimental results of water-injection tests at stresses
above (test Plw17) and below (tests Plw19 and Plw19a),
respectively, the hydrostatic plastic yield stress of the fully
10.34 saturated material. The load-unload cycles for tests Plw19
6.46 and Plw19a show indeed an elastic behavior both before
and after water injection. In the figure, the results from
the hydrostatic tests without water injection on specimens
with saturation 0.013, 0.019 and 1 are superimposed for
comparison. During water injection the specimens experience
initially a small dilation, followed by compaction. The total
final compaction approaches the compaction obtained in the
fully saturated specimen at the corresponding stress level. In
the specimen water-injected at a stress level above the plastic
yield stress of the fully saturated material, the additional
water-injection-related compaction is significant, due to pore
collapse during water injection. The initial dilation may be
related to the elimination of the capillary suction during
water injection.
 E Papamichos / Constitutive Laws for Geomaterials 767

-70 differentiation of Equation (23) yields a relation between the

Plw10 S = 0.013
effective stress ij rate and the elastic strain rate ije :
-60 Plw22 S = 0.019 e
dCijkl
Hydrostatic stress (MPa)

Plw31 S = 1
-50 Plw17 wat. inj. ij = Cijkl
e
ekl + Dije S , Dije = ekl (24)
Plw19a wat. inj. dS
Plw19 wat. inj.
-40
The plastic strain rate ijp is obtained by a nonassociated
-30 flow rule as:
Q
-20 ijp = (25)
ij
-10

0 ( )
where Q = Q ij , , S is the plastic potential function, and
0 -0.005 -0.01 -0.015 -0.02 -0.025 -0.03
is a scalar function. From the consistency condition:
Volumetric strain

Figure 10 F F F
F = ij + + S=0 (26)
Water-injection tests on Pietra Leccese chalk at hydrostatic ij S
stress 10, 20 and 40 MPa. The results from the hydrostatic
compression tests without water injection are included for
comparison. and using Equations (22), (24) and (25), can be solved as:

1 F e 1 F F e
= Cijkl kl + + Dij S
h ij h S ij
2.2 Model Description (27)
F e Q F
In partially saturated media, the generalized effective stress h= Cijkl
ij kl
ij is given as:

(
ij = ij + pgas ij S pgas pw ij ) (20) For plastic loading, is a nonnegative factor and h is
always positive. It then follows from Equation (27) that the
where ij is the total stress, S is the water saturation, pgas and plastic loading criterion can be expressed in terms of the
pw the pore gas and pore water pressures, respectively, and switch function 1 defined as:
ij the Kronecker delta. By referring all pressures to the 1 for F = 0 and
atmospheric pressure, pgas is zero when equal to the F e F F e
atmospheric, and Equation (20) can be written as: Cijkl kl + + Dij S > 0
ij S ij
1 =
ij = ij Spcp ij (21) (28)
0 for F < 0 or F = 0 and
F e F F e
where the pore-water capillary suction pcp, defined as Cijkl + + Dij S 0
ij S ij
pcp = pgas pw , is a function of the saturation S, that is
pcp = pcp (S). Plastic yield is associated with an isotropic yield
surface F, the size of which depends on the accumulated The plastic strain rate can be calculated from the flow rule
plastic strain parameter and the saturation S, that is Equation (25), and Equation (27) as:
( )
F ij , , S = 0 . The total strain rate ij is decomposed into
1 Q F e
two parts: ijp = Cstkl kl
h ij st
ij = ije + ijp (22) (29)
1 Q F F e
+ + Dkl S
The elastic strain ij is related to the effective stress ij h ij S kl
by a linear elastic law as:
ij = Cijkl
e
ekl (23) The effective stress rate ij can be determined from
Equation (24), using Equations (22) and (29) as:
e
where the elastic stiffness tensor Cijkl is assumed to be a
e
function of the saturation S, that is Cijkl = Cijkl
e
( S) . Time ij = Cijkl
ep
kl + Dijep S (30)
768 Oil & Gas Science and Technology Rev. IFP, Vol. 54 (1999), No. 6

ep
where the tangent elasto-plastic stiffness Cijkl and saturation conceptually replaces the sinus of the friction angle , and
ep
Dij tensors are given as: that the modified yield surface has been augmented with a
square root term for the mean pressure p and a new
1 e Q F e parameter pc related to the pressure cap. Figure 11 compares
ep
Cijkl = Cijkl
e
Cijmn Cstkl
h mn st the Mohr-Coulomb and the modified Mohr-Coulomb yield
(31)
surfaces in the -p plane for = 0.6, q = 5 MPa, pc = 10 MPa,
1 e Q F F e
D =D
ep
Cijkle
+ Dst and conditions of conventional triaxial compression where
kl S st
ij ij
h = /6.
Finally, the total stress rate ij can be obtained by time
differentiation of Equation (21) and substitution of ij from
Equation (30), as: 12

ij = Cijkl
ep
kl + Pij S, Pij =
(
d Spcp ) + Dijep (32)
Mohr-Coulomb

ij
dS 8

(MPa)
The employed elasticity is isotropic and linear with stress, Mohr-Coulomb
w/pressure cap
ep
in which case the elastic stiffness Cijkl and saturation Dije
tensors are given by the expressions: 4

2
e
Cijkl = G( S ) ij kl + ik jl + il jk
1 2 (33) 0
1 dG -pc q
Dij =
e
ij -12 -8 -4 0 4
G dS p (MPa)

where G is the elastic shear modulus and the Poisson's Figure 11

ratio. Based on the experimental data for the Pietra Leccese Illustration in the -p plane of the original Mohr-Coulomb
chalk, the dependency of the elasticity on the saturation S is and the Mohr-Coulomb with a pressure cap yield surfaces.
reflected only in the shear modulus G, which is assumed as a
function of the saturation.
The employed yield surface is a modified Mohr-Coulomb
yield function to include a pressure cap. It is described by the The corresponding plastic potential function to Equa-
expression: tion (34) is:

( )
F ij , , q, pc , S = cos

sin ( )
Q ij , , q, pc , S = cos
3
sin

3 (34) (36)
p + pc
( q p)
p + pc
=0 (q p)
q + pc q + pc

where the parameter replaces the sinus of the dilatancy angle

where three invariants, termed the isotropic (mean) effective
stress p, the shear stress intensity , and the Lode angle , in the Mohr-Coulomb model. The hardening parameters ,
have been introduced for the effective stress ij : are expressed as functions of the accumulated plastic shear
strain gp, that is = (gp), = (gp). The parameter q is
1 1 3 3 J3 (35) expressed as a function of the saturation S, while the
p= I1 , = J2 , = a sin 2 J3 2 parameter pc as a function of the saturation and the
3 3 2
accumulated plastic volumetric strain vp, that is q = q (S),
with I1 the first invariant of ij , and J2, J3 the second and pc = pc (S, vp). The plastic shear gp and volumetric vp strains
third invariants of the deviatoric stress Sij, respectively. In are defined as:
Equation (34) the hardening/softening parameters q, pc and
represent the size of the yield surface. The parameters q and
g p = gp dt , gp = 2eijp eijp
(37)
pc are identified as the intercepts of the yield surface F with v p
= v dt ,
p
vp = kkp
the positive (tension cut-off) and negative (pressure cap)
p-axis, respectively. A comparison of Equation (34) with the
p
Mohr-Coulomb yield function shows that the parameter where eij is the plastic deviatoric strain.
 E Papamichos / Constitutive Laws for Geomaterials 769

2.3 Calculation Examples -70

-65 Test Plw10: S = 0.013
The calibration procedure and the model parameters for the -60 Test Plw31: S = 1

Hydrostatic stress (MPa)

-55 Test Plw17: water injection
Pietra Leccese chalk can be found in Papamichos et al.
-50 Model
(1997). The hydrostatic compression tests on the specimens -45
with saturation S = 0.013, 0.019, 0.047, and 1 were back- -40
analyzed with the developed model and the results are pres- -35
-30
ented and compared with the experimental data in Figure 12. -25
The simulations compare well with the experimental data, -20
showing that the model can simulate properly the saturation- -15
-10
dependent elastic stiffness during the initial part of the test, -5
the saturation-dependent hydrostatic stress at plastic yield, 0
and reasonably well the pore collapse behavior of the Pietra 0 -0.01 -0.02 -0.03 -0.04 -0.05
Leccese chalk. The back analysis of the water-injection test Volumetric strain
at 40 MPa and the comparison with the experimental data are Figure 13
presented in Figure 13. The specimen enters the plastic
Theoretical simulations and experimental results of
regime of deformation and experiences pore collapse during hydrostatic compression tests with water injection at
the water-injection phase. The simulation compares well with hydrostatic compressive stress 40 MPa.
the experimental data, and the model can simulate properly
the saturation-dependent elastic stiffness during the initial
loading part of the test, the additional compaction during the
water-injection phase at both the elastic and plastic regimes,
and the elastic stiffness during the unloading part of the test. 3 SANDSTONE MODEL FOR CORE DAMAGE
In the model, the specimen straining during water injection
comprises of three parts: Core damage is a permanent alteration of the rock properties
a dilatant part due to the elimination of the capillary as a result of drill-out and retrieval of the core from the in situ
suction, which essentially results in a reduction of the environment to the surface. For reservoir rocks, stress release
effective compressive stress; during coring is thought to be a major core damage
a compactive elastic part due to the reduction of the elastic mechanism. Holt et al. (1998) have investigated core damage
stiffness with increasing water saturation; in reservoir sandstones using a synthetic sandstone,
a compactive plastic part due to plastification from the soft- analogous to the reservoir rock with respect to rock
ening of the pressure cap with increasing water saturation. mechanical and petrophysical properties and texture. The
synthetic sandstone is created under in situ stress conditions
to simulate the diagenesis process. Comparative tests are then
performed to simulate virgin compaction behavior, and the
Data S = 0.013 behavior of an unloaded and reloaded simulated core.
-180 Data S = 0.019
Data S = 0.047 The stress paths followed in the virgin compaction tests
-160 Data S = 1
Model and the simulated core compaction tests are shown in
Hydrostatic stress (MPa)

-140 Figure 14. In the virgin compaction tests, sand was loaded to
-120 the in situ stress conditions where cementation took place.
-100 After cementation, the specimen was compacted under
oedometric K0 conditions. In the simulated core compaction
-80
tests, after cementation at in situ stress conditions, the
-60 specimen was unloaded to simulate the coring process.
-40 Subsequently the specimen was loaded to the in situ stress
-20 conditions and an oedometric K0 compaction test was
0 performed to simulate a test on a cored specimen. Figure 15
0 -0.05 -0.1 -0.15 -0.2 shows a comparison between the virgin compaction and the
Volumetric strain simulated core compaction for this sandstone. Figure 15a
Figure 12 shows only the K0 part of the tests. The virgin compaction
Theoretical simulations and experimental results of
curves are nonlinear with a higher initial stiffness. The
hydrostatic compression tests on specimens with various stiffness during virgin compaction matches the stiffness of
water saturations S. the core compaction above 45 MPa. Thus, if core data are
used directly for reservoir rock compaction estimation, the
770 Oil & Gas Science and Technology Rev. IFP, Vol. 54 (1999), No. 6

Virgin uniaxial compaction (K0) Uniaxial compaction (K0)

of core

Axial and radial stresses

Axial and radial stresses

Axial stress Forming

Forming
Coring simulation Axial stress

Radial stress
Radial stress

b Time
a Time
Figure 14
Stress paths in compaction tests. a: virgin K0 compaction, and b: coring simulation, reloading to in situ conditions and K0 compaction (Holt et
al., 1998).

60 70
Virgin compaction Simulated core compaction
Virgin K0 compaction 60

50
Axial stress (MPa)

50
Axial stress (MPa)

40
A Gron01
30
A Gron03
40 A Gron14 20
Simulated B1s Gron05
core compaction B1s Gron04 10
B1f Gron15
30 0
0 2 4 6 8 -0.02 -0.015 -0.01 -0.005 0 0.005 0.01
a Axial strain (millistrain) b Axial strain

Figure 15
Compaction behavior of virgin versus simulated core of a synthetic reservoir sandstone (Holt et al., 1998). a: K0 phase of the tests, and b: all
tests.

initial compaction will be overestimated. In Figure 15b, the that the in situ stress conditions lie on it. During virgin K0
unloading during the coring simulation is also included. The loading, the sand matrix is further loaded and the yield
two specimens behave exactly the same above an axial stress surface extends further such that the current stress conditions
of 45 MPa. lie always on the yield surface. This is illustrated schematically
The test results show that during coring, unrecoverable in Figure16.
damage takes place with large plastic strains. This can be The cement at in situ stresses is not loaded and thus during
modeled with an elastoplastic model with a pressure cap to either coring or virgin K0 loading, it loads, and after an elastic
model correctly the behavior under uniaxial compaction. A region it yields. The yield surface for the cement is centered
way to describe the core damage process is a model with two at the in situ stress conditions and extends such that it gives
yield surfaces. Both yield surfaces have a pressure cap. Such an initial elastic region upon reduction of the stresses for
models are referred to as overlay or mechanical sublayer coring simulation or increase of the stresses for virgin K0
models. The one yield surface is for the sand matrix and the loading. Yielding leads to the destruction of the cement
other for the cement. The sand matrix at in situ stresses is cohesion, such that the cement yield surface reduces in size
loaded and thus, during coring, it simply unloads. The yield during stress changes, which can be either due to coring or
surface for sand is centered at zero stresses and extends such virgin K0 loading. Upon reloading to in situ stresses,
 E Papamichos / Constitutive Laws for Geomaterials 771

F for sand matrix 16 stresses. The internal stresses result from the fact that the two
F for cement K0 loading stress path materials, the sand matrix and the cement, act in opposite
14
Coring stress path ways in the case of unloading, that is coring. Internally

Deviatoric stress (MPa)

12 viewed the compression in the sand matrix is counter
10 balanced by tension in the cement, resulting in zero external
stress.
8
Figure 17 shows a simulation of virgin K0 compaction and
6 coring K0 compaction. The model results compare favorable
4 with the experimental results in Figure 15a showing the
potential of the model to predict core damage in reservoir
2
sandstones.
0
-30 -25 -20 -15 -10 -5 0 5
Mean stress p (MPa) CONCLUSIONS
Figure 16
The presented modeling examples show that the behavior of
Yield surfaces F for the sand matrix and the cement and geomaterials is often too complex to generalize the models.
stress paths during virgin and core compaction.
Enough testing material is often not available for a
calibration of a general constitutive model at a reasonable
cost. General models lead also to cumbersome computations,
-65 which often obscure the physics of the problem at hand. The
Coring + K0 role of the engineer becomes thus important, as he is required
-60
Direct K0 to identify and model first-order effects.
Axial stress (MPa)

-55

-50 REFERENCES

-45 Chen, W.F. and Han, D.J. (1988) Plasticity for Structural
Engineers, Springer-Verlag, 173.
-40 Holt, R.M., Brignoli, M., Kenter, C.J., Meij, R. and Schutjens,
P.M.T.M (1998) From Core Compaction to Reservoir Compaction:
-35 Correction for Core Damage Effects. SPE/ISRM 47263. Proc.
Eurock 98, 1, 311-320.
-30 Papamichos, E., Brignoli, M. and Santarelli, F.J. (1997) An
0 -0.001 -0.002 -0.003 -0.004 -0.005 -0.006 -0.007 -0.008 -0.009 Experimental and Theoretical Study of a Partially-Saturated
Axial strain Collapsible Rock. Mech. Cohes.-Frict. Mater., 2, 3, 251-278.
Papamichos, E., Tronvoll, J., Skjrstein, A., Unander, T.E.,
Figure 17 Vardoulakis, I. and Sulem, J. (1996) The Effect of Plane-Strain
Simulation of virgin K0 compaction and coring K0 compaction. and Isotropic Loading in Hollow-Cylinder Strength. Proc.
Eurock 96, Torino, Barla, G. (ed.), 1, Balkema, Rotterdam, 197-
204.
Papamichos, E., Tronvoll, J., Vardoulakis, I., Labuz, J.F.,
nonlinear but elastic behavior is observed. After that the Skjrstein, A., Unander, T.E. and Sulem, J. (1999) Constitutive
Testing of Red Wildmoor Sandstone. Mech. Cohes.-Frict. Mater., 4.
material continues to yield since the cement has been
Sulem, J., Vardoulakis, I., Papamichos, E., Oulahna, A. and
destructed. Essentially destruction of the cement occurs Tronvoll, J. (1999) Elasto-Plastic Behaviour of Red Wildmoor
either by loading or by unloading from in situ conditions, Sandstone. Mech. Cohes.-Frict. Mater., 4.
such that after the cement destruction the two materials Truesdell, C. and Noll, W. (1965) The Non-Linear Field Theories
behave exactly the same. of Mechanics. In Encyclopedia of Physics, Flgge, S. (ed.), III/3,
Sect. 3, Springer-Verlag.
The proposed two-material model can also be viewed as a
model with internal, often referred to also as residual, Final manuscript received in July 1999